Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[KBL][ffmpeg-vaapi][vp8] decoding fails #317

Closed
fulinjie opened this issue Sep 14, 2018 · 5 comments
Closed

[KBL][ffmpeg-vaapi][vp8] decoding fails #317

fulinjie opened this issue Sep 14, 2018 · 5 comments
Assignees
Labels
verifying PR: fix ready and verifying with build/test

Comments

@fulinjie
Copy link
Contributor

cmd line:
ffmpeg -hwaccel vaapi -vaapi_device /dev/dri/renderD128 -v verbose -y -i ./vp80-02-inter-1418.ivf -f md5 ffmpeg_vaapi.md5

media driver:
commit id:
6ecb1d7

src file:
https://github.com/webmproject/vp8-test-vectors/blob/master/vp80-02-inter-1418.ivf

error message:
[vp8 @ 0x55a85df70d40] Failed to end picture decode issue: 23 (internal decoding error).
Error while decoding stream #0:0: Input/output error
[vp8 @ 0x55a85df73e40] Failed to end picture decode issue: 23 (internal decoding error).
[vp8 @ 0x55a85df76f40] Failed to end picture decode issue: 23 (internal decoding error).
Error while decoding stream #0:0: Input/output error
Last message repeated 1 times
[vp8 @ 0x55a85df603c0] Failed to end picture decode issue: 23 (internal decoding error).
Error while decoding stream #0:0: Input/output error
[vp8 @ 0x55a85df617c0] Failed to end picture decode issue: 23 (internal decoding error).
Error while decoding stream #0:0: Input/output error
[vp8 @ 0x55a85df64940] Failed to end picture decode issue: 23 (internal decoding error).
Error while decoding stream #0:0: Input/output error
Last message repeated 6 times
No more output streams to write to, finishing.
frame= 79 fps=0.0 q=-0.0 Lsize= 0kB time=00:00:03.60 bitrate= 0.1kbits/s speed= 49x
video:4629kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: unknown
Input file #0 (./vp80-02-inter-1418.ivf):
Input stream #0:0 (video): 108 packets read (109579 bytes); 79 frames decoded;
Total: 108 packets (109579 bytes) demuxed
Output file #0 (ffmpeg_vaapi.md5):
Output stream #0:0 (video): 79 frames encoded; 79 packets muxed (4740000 bytes);
Total: 79 packets (4740000 bytes) muxed

This file could be decoded successfully on i965 driver.

@XinfengZhang
Copy link
Contributor

please help to verify https://patchwork.ffmpeg.org/patch/9943/
thanks

@XinfengZhang XinfengZhang added the verifying PR: fix ready and verifying with build/test label Sep 16, 2018
@wangyan42164
Copy link
Contributor

I tried this patch in the latest ffmpeg on KBL. The error is gone with the source file.

@fulinjie
Copy link
Contributor Author

This patch works. Thanks.

@uartie
Copy link
Contributor

uartie commented Sep 18, 2018

works for me!

mypopydev added a commit to mypopydev/FFmpeg that referenced this issue Oct 17, 2018
Fix the issue: intel/media-driver#317

the root cause is update_dimensions call get_pixel_format will
trigger the hwaccel_uninit/hwaccel_init , in current context,
there are 3 situations in the update_dimensions():
1. First time calling. No matter single thread or multithread,
   get_pixel_format() should be called after dimensions were
   set;
2. Dimention changed at the runtime. Dimention need to be
   updated when macroblocks_base is already allocated,
   get_pixel_format() should be called to recreate new frames
   according to updated dimention;
3. Multithread first time calling. After decoder init, the
   other threads will call update_dimensions() at first time
   to allocate macroblocks_base and set dimensions.
   But get_pixel_format() is shouldn't be called due to low
   level frames and context are already created.

In this fix, we only call update_dimensions as need.

Signed-off-by: Wang, Shaofei <shaofei.wang@intel.com>
Reviewed-by: Jun, Zhao <jun.zhao@intel.com>
Reviewed-by: Haihao Xiang <haihao.xiang@intel.com>
xhaihao pushed a commit to xhaihao/FFmpeg that referenced this issue Jul 14, 2020
Fix the issue: intel/media-driver#317

the root cause is update_dimensions will be called multple times
when decoder thread number is not only 1, but update_dimensions
call get_pixel_format in each decode thread will trigger the
hwaccel_uninit/hwaccel_init more than once. But only one hwaccel
should be shared with all decode threads.
in current context,
there are 3 situations in the update_dimensions():
1. First time calling. No matter single thread or multithread,
   get_pixel_format() should be called after dimensions were
   set;
2. Dimention changed at the runtime. Dimention need to be
   updated when macroblocks_base is already allocated,
   get_pixel_format() should be called to recreate new frames
   according to updated dimention;
3. Multithread first time calling. After decoder init, the
   other threads will call update_dimensions() at first time
   to allocate macroblocks_base and set dimensions.
   But get_pixel_format() is shouldn't be called due to low
   level frames and context are already created.

In this fix, we only call update_dimensions as need.

Signed-off-by: Wang, Shaofei <shaofei.wang@intel.com>
Reviewed-by: Jun, Zhao <jun.zhao@intel.com>
Reviewed-by: Haihao Xiang <haihao.xiang@intel.com>
xhaihao pushed a commit to xhaihao/FFmpeg that referenced this issue Jul 23, 2020
Fix the issue: intel/media-driver#317

the root cause is update_dimensions will be called multple times
when decoder thread number is not only 1, but update_dimensions
call get_pixel_format in each decode thread will trigger the
hwaccel_uninit/hwaccel_init more than once. But only one hwaccel
should be shared with all decode threads.
in current context,
there are 3 situations in the update_dimensions():
1. First time calling. No matter single thread or multithread,
   get_pixel_format() should be called after dimensions were
   set;
2. Dimention changed at the runtime. Dimention need to be
   updated when macroblocks_base is already allocated,
   get_pixel_format() should be called to recreate new frames
   according to updated dimention;
3. Multithread first time calling. After decoder init, the
   other threads will call update_dimensions() at first time
   to allocate macroblocks_base and set dimensions.
   But get_pixel_format() is shouldn't be called due to low
   level frames and context are already created.

In this fix, we only call update_dimensions as need.

Signed-off-by: Wang, Shaofei <shaofei.wang@intel.com>
Reviewed-by: Jun, Zhao <jun.zhao@intel.com>
Reviewed-by: Haihao Xiang <haihao.xiang@intel.com>
xhaihao pushed a commit to xhaihao/FFmpeg that referenced this issue Jul 27, 2020
Fix the issue: intel/media-driver#317

the root cause is update_dimensions will be called multple times
when decoder thread number is not only 1, but update_dimensions
call get_pixel_format in each decode thread will trigger the
hwaccel_uninit/hwaccel_init more than once. But only one hwaccel
should be shared with all decode threads.
in current context,
there are 3 situations in the update_dimensions():
1. First time calling. No matter single thread or multithread,
   get_pixel_format() should be called after dimensions were
   set;
2. Dimention changed at the runtime. Dimention need to be
   updated when macroblocks_base is already allocated,
   get_pixel_format() should be called to recreate new frames
   according to updated dimention;
3. Multithread first time calling. After decoder init, the
   other threads will call update_dimensions() at first time
   to allocate macroblocks_base and set dimensions.
   But get_pixel_format() is shouldn't be called due to low
   level frames and context are already created.

In this fix, we only call update_dimensions as need.

Signed-off-by: Wang, Shaofei <shaofei.wang@intel.com>
Reviewed-by: Jun, Zhao <jun.zhao@intel.com>
Reviewed-by: Haihao Xiang <haihao.xiang@intel.com>
xhaihao pushed a commit to xhaihao/FFmpeg that referenced this issue Jul 28, 2020
Fix the issue: intel/media-driver#317

the root cause is update_dimensions will be called multple times
when decoder thread number is not only 1, but update_dimensions
call get_pixel_format in each decode thread will trigger the
hwaccel_uninit/hwaccel_init more than once. But only one hwaccel
should be shared with all decode threads.
in current context,
there are 3 situations in the update_dimensions():
1. First time calling. No matter single thread or multithread,
   get_pixel_format() should be called after dimensions were
   set;
2. Dimention changed at the runtime. Dimention need to be
   updated when macroblocks_base is already allocated,
   get_pixel_format() should be called to recreate new frames
   according to updated dimention;
3. Multithread first time calling. After decoder init, the
   other threads will call update_dimensions() at first time
   to allocate macroblocks_base and set dimensions.
   But get_pixel_format() is shouldn't be called due to low
   level frames and context are already created.

In this fix, we only call update_dimensions as need.

Signed-off-by: Wang, Shaofei <shaofei.wang@intel.com>
Reviewed-by: Jun, Zhao <jun.zhao@intel.com>
Reviewed-by: Haihao Xiang <haihao.xiang@intel.com>
xhaihao pushed a commit to xhaihao/FFmpeg that referenced this issue Jul 29, 2020
Fix the issue: intel/media-driver#317

the root cause is update_dimensions will be called multple times
when decoder thread number is not only 1, but update_dimensions
call get_pixel_format in each decode thread will trigger the
hwaccel_uninit/hwaccel_init more than once. But only one hwaccel
should be shared with all decode threads.
in current context,
there are 3 situations in the update_dimensions():
1. First time calling. No matter single thread or multithread,
   get_pixel_format() should be called after dimensions were
   set;
2. Dimention changed at the runtime. Dimention need to be
   updated when macroblocks_base is already allocated,
   get_pixel_format() should be called to recreate new frames
   according to updated dimention;
3. Multithread first time calling. After decoder init, the
   other threads will call update_dimensions() at first time
   to allocate macroblocks_base and set dimensions.
   But get_pixel_format() is shouldn't be called due to low
   level frames and context are already created.

In this fix, we only call update_dimensions as need.

Signed-off-by: Wang, Shaofei <shaofei.wang@intel.com>
Reviewed-by: Jun, Zhao <jun.zhao@intel.com>
Reviewed-by: Haihao Xiang <haihao.xiang@intel.com>
xhaihao pushed a commit to xhaihao/FFmpeg that referenced this issue Aug 18, 2020
Fix the issue: intel/media-driver#317

the root cause is update_dimensions will be called multple times
when decoder thread number is not only 1, but update_dimensions
call get_pixel_format in each decode thread will trigger the
hwaccel_uninit/hwaccel_init more than once. But only one hwaccel
should be shared with all decode threads.
in current context,
there are 3 situations in the update_dimensions():
1. First time calling. No matter single thread or multithread,
   get_pixel_format() should be called after dimensions were
   set;
2. Dimention changed at the runtime. Dimention need to be
   updated when macroblocks_base is already allocated,
   get_pixel_format() should be called to recreate new frames
   according to updated dimention;
3. Multithread first time calling. After decoder init, the
   other threads will call update_dimensions() at first time
   to allocate macroblocks_base and set dimensions.
   But get_pixel_format() is shouldn't be called due to low
   level frames and context are already created.

In this fix, we only call update_dimensions as need.

Signed-off-by: Wang, Shaofei <shaofei.wang@intel.com>
Reviewed-by: Jun, Zhao <jun.zhao@intel.com>
Reviewed-by: Haihao Xiang <haihao.xiang@intel.com>
xhaihao pushed a commit to xhaihao/FFmpeg that referenced this issue Sep 8, 2020
Fix the issue: intel/media-driver#317

the root cause is update_dimensions will be called multple times
when decoder thread number is not only 1, but update_dimensions
call get_pixel_format in each decode thread will trigger the
hwaccel_uninit/hwaccel_init more than once. But only one hwaccel
should be shared with all decode threads.
in current context,
there are 3 situations in the update_dimensions():
1. First time calling. No matter single thread or multithread,
   get_pixel_format() should be called after dimensions were
   set;
2. Dimention changed at the runtime. Dimention need to be
   updated when macroblocks_base is already allocated,
   get_pixel_format() should be called to recreate new frames
   according to updated dimention;
3. Multithread first time calling. After decoder init, the
   other threads will call update_dimensions() at first time
   to allocate macroblocks_base and set dimensions.
   But get_pixel_format() is shouldn't be called due to low
   level frames and context are already created.

In this fix, we only call update_dimensions as need.

Signed-off-by: Wang, Shaofei <shaofei.wang@intel.com>
Reviewed-by: Jun, Zhao <jun.zhao@intel.com>
Reviewed-by: Haihao Xiang <haihao.xiang@intel.com>
xhaihao pushed a commit to xhaihao/FFmpeg that referenced this issue Sep 17, 2020
Fix the issue: intel/media-driver#317

the root cause is update_dimensions will be called multple times
when decoder thread number is not only 1, but update_dimensions
call get_pixel_format in each decode thread will trigger the
hwaccel_uninit/hwaccel_init more than once. But only one hwaccel
should be shared with all decode threads.
in current context,
there are 3 situations in the update_dimensions():
1. First time calling. No matter single thread or multithread,
   get_pixel_format() should be called after dimensions were
   set;
2. Dimention changed at the runtime. Dimention need to be
   updated when macroblocks_base is already allocated,
   get_pixel_format() should be called to recreate new frames
   according to updated dimention;
3. Multithread first time calling. After decoder init, the
   other threads will call update_dimensions() at first time
   to allocate macroblocks_base and set dimensions.
   But get_pixel_format() is shouldn't be called due to low
   level frames and context are already created.

In this fix, we only call update_dimensions as need.

Signed-off-by: Wang, Shaofei <shaofei.wang@intel.com>
Reviewed-by: Jun, Zhao <jun.zhao@intel.com>
Reviewed-by: Haihao Xiang <haihao.xiang@intel.com>
@uartie
Copy link
Contributor

uartie commented Sep 21, 2020

FFMpeg patch never got merged to upstream... see #1028

TimothyGu pushed a commit to FFmpeg/FFmpeg that referenced this issue Sep 23, 2020
Fix the issue: intel/media-driver#317

the root cause is update_dimensions will be called multple times
when decoder thread number is not only 1, but update_dimensions
call get_pixel_format in each decode thread will trigger the
hwaccel_uninit/hwaccel_init more than once. But only one hwaccel
should be shared with all decode threads.
in current context,
there are 3 situations in the update_dimensions():
1. First time calling. No matter single thread or multithread,
   get_pixel_format() should be called after dimensions were
   set;
2. Dimention changed at the runtime. Dimention need to be
   updated when macroblocks_base is already allocated,
   get_pixel_format() should be called to recreate new frames
   according to updated dimension;
3. Multithread first time calling. After decoder init, the
   other threads will call update_dimensions() at first time
   to allocate macroblocks_base and set dimensions.
   But get_pixel_format() is shouldn't be called due to low
   level frames and context are already created.

In this fix, we only call update_dimensions as need.

Signed-off-by: Wang, Shaofei <shaofei.wang@intel.com>
Reviewed-by: Jun, Zhao <jun.zhao@intel.com>
Reviewed-by: Haihao Xiang <haihao.xiang@intel.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
Kwiboo pushed a commit to Kwiboo/FFmpeg that referenced this issue Nov 30, 2020
Fix the issue: intel/media-driver#317

the root cause is update_dimensions will be called multple times
when decoder thread number is not only 1, but update_dimensions
call get_pixel_format in each decode thread will trigger the
hwaccel_uninit/hwaccel_init more than once. But only one hwaccel
should be shared with all decode threads.
in current context,
there are 3 situations in the update_dimensions():
1. First time calling. No matter single thread or multithread,
   get_pixel_format() should be called after dimensions were
   set;
2. Dimention changed at the runtime. Dimention need to be
   updated when macroblocks_base is already allocated,
   get_pixel_format() should be called to recreate new frames
   according to updated dimension;
3. Multithread first time calling. After decoder init, the
   other threads will call update_dimensions() at first time
   to allocate macroblocks_base and set dimensions.
   But get_pixel_format() is shouldn't be called due to low
   level frames and context are already created.

In this fix, we only call update_dimensions as need.

Signed-off-by: Wang, Shaofei <shaofei.wang@intel.com>
Reviewed-by: Jun, Zhao <jun.zhao@intel.com>
Reviewed-by: Haihao Xiang <haihao.xiang@intel.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit eb6bb8f)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
verifying PR: fix ready and verifying with build/test
Projects
None yet
Development

No branches or pull requests

5 participants